Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
AI Driven InsurerEmerging Risk & Opportunity in a Digital WorldA View from the Startup, DataRobot Under the
Spotlight by Satadru Sengupta and Sophie Roberts
Dubai World Insurance Congress | February 2018
Satadru SenguptaGM Insurance, DataRobot Inc.
Confidential | Copyright © DataRobot, Inc. | All Rights Reserved
The world’s most advanced Enterprise Machine Learning Automation platform
2012Founded, HQ in Boston, MA
$124MIn funding
400,000,000+Models built on DataRobot Cloud
180+Data Scientists & Engineers (of 300+)
4#1 ranked Data Scientists
50+Top 3 finishes
INSURANCE FINTECH HEALTHCARE MARKETING BANKING MANY MORE
Fresh Capital From Big PlayersBlurred line between Silicon Valley & Wall St.
Trickle down effect on medium and small players
"New technology is transforming the way we work, and it is allowing the competition to do
better than what we can.
The strange thing is we know the urgency, and yet there is inertia."
Inga Beale, The Lloyds of London, DWIC 2017
1. Elastic Cloud Computing (Cloud)
2. Internet of Things (IoT)
3. Artificial Intelligence (AI)
Elements of Digital Transformation
Confidential | Copyright © DataRobot, Inc. | All Rights Reserved
AI & Machine Learning
monetizes the data
“Whatever else it produces, an organization is a factory that produces judgments and decisions.”
Daniel Kahneman, Thinking Fast & Slow
Machine Learning
Deep Learning
Better Decision, Faster Decision.
Machine doesn’t fall for biases.
Augmentation as opposed to “full autonomy”
Machine Learning
Deep Learning
AI: computer systems able to perform tasks that ordinarily require human intelligence
Machine Learning
Deep Learning
AI: computer systems able to perform tasks that ordinarily require human intelligence
Powered by boring stuff
Powered by deep learning
Powered by ML (non-DL)
Confidential. Copyright © DataRobot, Inc. - All Rights Reserved
How Machine Learning Is Used In Insurance
Historical Examples to Learn From
Machine Learning
Algorithm+ =
New
Dat
a Good Risk
Poor Risk
Average Risk
What are the risks an insurance company cares about?● Marketing & Distribution: acquisition, retention (short-term policies, e.g. as in P&C), lapse
(long-term policies, e.g. in life), submission prioritization, product recommendation
● Underwriting: underwriting risk (# claims, $ severity, mortality, health), agent mis-conduct, risk factors mis-reporting (roof age, insured age, smoking status, mileage, diabetes, family history)
● Claims: fraud, litigation, subro/ recovery, severity, duration
● Others: audit, IoT, loss prevention
Confidential. Copyright © DataRobot, Inc. - All Rights Reserved
P&C Use Case: Claims Management● Manage claims proactively to avoid fraud, contain severity of claims, and reduce operational
cost
● Predict the claim outcome based on the information available at the First Notice of Loss (FNOL) and accordingly triage the claim to the right adjustor and/ or SIU
○ Predict the likelihood of claim being a fraud claim and assign high risk claims it to SIU for manual review
○ Predict complexity of the claim (if complex and long duration then propose settlement and assign to specialized adjustor):
■ Size of the claim■ Duration of the claim■ Likelihood of litigation
○ Predict likelihood of subrogation opportunity
● How we operationalize? Claims modeling requires transparency that will help claim adjustors and fraud investigator understand what are the drivers of the predictive model and predictions.
● Economic value: increased operational efficiency, loss avoidance, loss containment
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
PREREQUISITES
1. Knowledge of the overall & specific problem2. Knowledge of the data3. Ability to write code to gather data4. Ability to write code to explore/inspect data5. Ability to write code to manipulate data6. Ability to write code to extract actionable intel7. Ability to write code to build models8. Ability to write code to implement models 9. Foundational statistics
10. Internals of algorithms 11. Practical knowledge and experience12. Knowing how to interpret and explain models
Requirement without AI & Robotic Process Automation
Math &Stats
DomainExpertise
ProgrammingSkills
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
Automated Machine Learning
AI (artificial intelligence) builds robotic
process automation for predictive
analytics - that focuses on accuracy,
transparency, and ease of use.
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
PREREQUISITES
1. Knowledge of the overall & specific problem2. Knowledge of the data
Requirement with AI & Robotic Process Automation
Math &Stats
DomainExpertise
ProgrammingSkills
AUTOMATED
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
Predictive Analytics Challenge
Data Scientists
Domain Expertise
Your Employees
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
The “Failed” Approach: Data Scientist Focused
Data Scientists
Domain Expertise
Your Employees
https://hbr.org/2016/12/why-youre-not-getting-value-from-your-data-science
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
Large Scale Delivery Model: Process Automation Focused
Data Scientists
Domain Expertise
Your Employees
MIT Tech Review Article
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
Accuracy
One or Two Algorithms.
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
Transparency
Only Selected Few Can Explain
Anyone Can Explain
Confidential. Copyright © DataRobot, Inc. and Cloudera, Inc. - All Rights Reserved
Ease of Use
Artisanal: Only Selected Few Can Use
Anyone Can Use
THANK YOU
[email protected]/insurance
VISIT US AT THE BOOTH TO SEE HOW AUTOMATED MACHINE LEARNING WORKS